A Weighted Tile-based Approach for Viewport Adaptive 360° Video Streaming

Autor: Gabriel-Miro Muntean, Abid Yaqoob
Rok vydání: 2020
Předmět:
Zdroj: BMSB
DOI: 10.1109/bmsb49480.2020.9379517
Popis: Recently, affordable virtual reality (VR) technology has experienced increased attention as it enables innovative and immersive 360° media applications. However, the expected immersive rich media streaming experience is hindered by excessive bandwidth requirements of high-resolution 360° content. It is essential to devise technologies that enable persistent delivery of such bandwidth-hungry applications over the best-effort packet networks. Tile-based adaptive streaming is one of such techniques that support differentiated streaming of quality-variable tiles depending on their position within the viewport. This paper proposes a weighted tile-based approach for 360° video streaming. First, a content-agnostic viewport prediction approach is proposed. Second, a novel viewport and bitrate adaptation heuristic solution which dynamically selects the coverage of the viewport and its marginal regions and allocates weights to the tiles in these regions is introduced. Finally, suitable video bitrates for each tile based on its weight and available network throughput is allocated. The solution is tested and assessed in comparison with alternative approaches resulting in improved viewport quality and lower bandwidth utilization.
Databáze: OpenAIRE